Measuring Gait of Total Knee Arthroplasty Patients Pre-Op and Post-Op Using a Smartphone

Motion capture technology is a tool often used in orthopedic applications, including measuring gait dysfunctions. However, gait analysis research focused on total knee arthroplasty (TKA) patients is limited, especially using motion capture systems. This may be due to a multitude of factors, such as inaccessibility to gait labs to the general public, difficulty of travel post-surgery, and overall costs of care and maintenance of motion lab equipment. New emerging novel methods use wearable devices such as smartphones and inertial measurement devices (IMUs) to perform gait assessment and are becoming favorable due to their increasingly ubiquitous nature. One such method using smartphones is an app called OneStep, which uses the smartphone’s sensors and machine learning algorithms to measure gait parameters of walking trials. The thesis project presented here used the OneStep app to measure stride length, step length, step width, gait velocity, cadence, and double stance time from walking trials. This study was two-fold with pilot and in-clinic studies, in which trials were conducted with Cal Poly students and TKA patients, respectively. The purpose of the pilot experiments was to validate the reliability of the OneStep app against gait variables calculated from traditional motion analysis software by conducting walking trials with the OneStep app and motion analysis system in the Mobile Biomechanics Lab (MBL) and walking trials with the OneStep app in the building hallway. Gait algorithms were created in MATLAB software to calculate gait parameters from heel and sacrum marker motion data and validate OneStep trials in the MBL. Results of the pilot experiments indicated statistical similarities for stride length, left step length, right step length, step width, cadence, and gait velocity between methods (OneStep in the MBL vs. Cortex) and for step width and cadence between walking conditions (OneStep in the MBL vs. OneStep in the hallway). Low reliability was observed for step width between methods and between walking conditions (R = 0.013, R = 0.35). Scatterplots comparing gait variables between methods indicated good visual agreement for stride length, left step length, right step length, and gait velocity. Strong visual agreement was observed for cadence. Low agreement was observed for step width and double stance time. Scatterplots comparing gait variables between walking conditions indicated good visual agreement for stride length, left step length, right step length, gait velocity, and step width. Moderate visual agreement was observed for step width and double stance time. Results of the clinic experiments using the OneStep app indicated statistical differences in stride length, left and right step length, and gait velocity (p = 0.0010, p = 0.0087, p = 0.015, p = 0.0070) between pre-operative and 2 weeks post-operative appointments. Future work of this project is to continue monitoring the functional recovery of patients until they meet or exceed their pre-operative gait values, performing pilot experiments with individuals with gait abnormalities, and using force plate data to determine toe-off events for gait algorithms calculated in MATLAB.